Hi Ricky,

Yes, it is possible. I think this thread (and the file linked therein) is exactly what you need: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;c9a13f1e.1506

Also, this other one might be useful, as it includes other contrasts of potential interest: https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;dbf8f005.1503

If all 4 observations per subject have the same pairwise covariances, then they are exchangeable, and you can use randomise with the option "-e design.grp" to supply the exchangeability blocks (EB). In PALM the same is done with "-eb design.grp".

If the above assumption about the covariance within-subject isn't met, but the errors are symmetric around zero (typically they are for FMRI data), then you can use whole-block sign-flipping: in randomise use the options "-e design.grp --permuteBlocks -1", or in PALM the options "-eb design.grp -whole -ise".

More details about this are in the randomise paper.

Hope this helps.

All the best,

Anderson


On 5 March 2016 at 03:58, Ricky Savjani <[log in to unmask]> wrote:
Hi FSLers,

We were wondering if there is a way to run a 2x2 ANOVA with repeated measures for each subject for each factor and level in randomize or in PALM?

That is, we have 4 PEs for every subject organized like this:

Factor A
     Level 1
     Level 2

Factor B
     Level 1
     Level 2

Each subject has measures for all conditions, giving 4 PEs for each of our 67 subjects. We would like to ask is there a main effect of Factor A, for Factor B, and if there is an interaction effect. However, we know we need to account for the fact that that there are repeated measures.

In AFNI, we can do such a command with 3dANOVA3 as follows:

3dANOVA3 -type 4 -alevels 2 -blevels 2 -clevels 67

where -alevels 2 states there are two levels for factor A, -blevels 2 states that there are 2 levels for factor B, -c levels states that there 67 subjects for the subject factor, which is treated as a random effect:

the -type 4 states is used to denote:

A,B factors are fixed; C random;  AxBxC

We would like to run this non-parametrically with FSL in PALM (or Randomize). Is it possible to construct such a design and/or run a variant of this type of ANOVA?

Many thanks in advance!

-Ricky